New Home Sales Fell 41% in February 2009

WSJ: Sales of new homes rose in February for the first time in seven months, the Commerce Department reported Wednesday, another sign that the housing market is thawing

Bloomberg: Purchases of new homes in the U.S. unexpectedly rose in February from a record low as plummeting prices and cheaper mortgage rates lured some buyers. Sales increased 4.7 percent to an annual pace of 337,000 . . .

Marketwatch: The U.S. housing sector continues to see signs of improvement. The latest government data showed new home sales climbed in February for the first time in seven months, sending shares of home-building companies soaring.

>

The parade of the mathematically innumerate business writers continue to misread data. The latest evidence? New Home Sales.

After incorrectly reporting the Existing Home Sales, the mainstream media misread the Census department report of New Homes.

No, New Home Sales data did not improve. In fact, they were not only not positive, they were actually horrific. The year over year number was a terrible down 41%.  Sales from this same period a year ago have nearly been halved.

Why did the media report this as positive? If you only read the headline number, you saw a positive datapoint: February was plus 4.7% over January.

To get the the facts, you need to read below the headline. In the present case, it wasn’t the seasonality factor that was confusing, it was the “90-percent confidence intervals” — or as it is more commonly known, the margin of error.

From the Census Bureau:

Sales of new one-family houses in February 2009 were at a seasonally adjusted annual rate of 337,000, according to estimates released jointly today by the U.S. Census Bureau and the Department of Housing and Urban Development. This is 4.7 percent (±18.3%)* above the revised January rate of 322,000, but is 41.1 percent (±7.9%) below the February 2008 estimate of 572,000.

The median sales price of new houses sold in February 2009 was $200,900; the average sales price was $251,000. The seasonally adjusted estimate of new houses for sale at the end of February was 330,000. This represents a supply of 12.2 months at the current sales rate.

Note that the month over month data at 4.7% — plus or minus 18.3% — is statistically insignificant. (i.e., meaningless). The reported data does not inform us if sales improved month-over-month or not. It is a range, from down -13.6% to plus 23%. Since “zero” is part of that range, we can draw no conclusion. As the Census Department itself notes, “the change is not statistically significant; that is, it is uncertain whether there was an increase or decrease.”

The data does however, tell us that the year-over-year sales fell 41.1% plus or minus 7.9% gives us a range of -49% to -33.2%. The entire range is negative, therefore we can conclude sales fell year-over-year.

These are facts. This is data. This is how you interpret it. Most of the MSM reports (WSJ, Marketwatch, Bloomberg) were simply wrong.

Not nuanced, not shaded, but 2+2=5 wrong.

Let me remind that many of these folks incorrectly misinformed you that Housing wasn’t getting worse in 2006, 2007 and 2008 — just as Home sales and prices went into an historic freefall. Now, these same folks are misinforming you that Housing has turned around and is improving. That is simply unsupported by the data.

(And don’t even ask about television — they simply read the wrong news. Here is a life lesson for you: Never believe news people who read teleprompters. They have no idea what they are doing, they are reading what someone else wrote. When it comes to data interpretation, they are quite literally clueless. Rely on news readers to your personal financial detriment).

The bottom line: Learn to interpret data correctly. Avoid using the people who cannot do so as primary news sources.

>

New Home Sales Annual Sales Rate, Seasonal Adjusted

c25_curr1.gif

Chart via Census Department

>

Previously: A Closer Look at New Home Sales Data (October 2006) http://www.ritholtz.com/blog/2006/10/a-closer-look-at-new-home-sales-data/

April New Home Sales – Revisited (November 2005) http://www.ritholtz.com/blog/2005/11/new-home-sales-data-dont-rely-on-it-either/

New Home Sales Fall 42% (May 2008) http://www.ritholtz.com/blog/2008/05/new-home-sales-fall-42/

Sources: NEW RESIDENTIAL SALES IN FEBRUARY 2009 http://www.census.gov/briefrm/esbr/www/esbr051.html http://www.census.gov/newhomesales

New-Home Sales Rise 4.7% KELLY EVANS WSJ, MARCH 26, 2009 http://online.wsj.com/article/SB123798406285137541.html

New-Home Sales in U.S. Rose 4.7% to a 337,000 Pace Shobhana Chandra Bloomberg, March 25 2009 http://www.bloomberg.com/apps/news?pid=newsarchive&sid=a0YI6z_ZfAUE

Home-builder shares jump as February new home sales increase Dawn Wotapka Marketwatch, March 25, 2009 http://tinyurl.com/c3pcmc

Unexpected Increase in Factory Orders JACK HEALY NYT, March 25, 2009 http://www.nytimes.com/2009/03/26/business/economy/26econ.html

The New York Times came closest to getting it right

“In another report, the government said that new single-family home sales rose 4.7 percent in February, but that it was still the second-worst month on record, and was down more than 40 percent from February a year ago.”


Originally published at The Big Picture blog and reproduced here with the author’s permission.

24 Responses to "New Home Sales Fell 41% in February 2009"

  1. Anonymous   March 26, 2009 at 11:01 am

    Thank you very much for the analysis. Most of the people don’t have the inclination or the resources to dig into the numbers, such as the ones described by you (where are the editors?) So, you are doing a public service. Great!

  2. Nick Giudici   March 26, 2009 at 11:43 am

    I find it disingenuous to post a criticism of someone’s ability to interpret data and then have a clearly misleading graph displaying the same data.The y axis of your graph does not start at 0 thus making the drop even more precipitous. If the graph was generated with the home sales starting a 0 then the final data point would be nearly three times the hight thus making the percent difference in area obviously much smaller.

  3. Anonymous   March 26, 2009 at 12:17 pm

    Re: Nick GiudiciMaking the graph start at zero would not have made much of a difference, the slope of the drop is whats significant, it would be the same if it was shifted vertically.

  4. Anonymous   March 26, 2009 at 12:17 pm

    Re: Nick GiudiciMaking the graph start at zero would not have made much of a difference, the slope of the drop is whats significant, it would be the same if it was shifted vertically.

  5. Guest   March 26, 2009 at 12:59 pm

    You’re wrong to say the data is “meaningless.” If you assume a normal bellcurve, then the data gives you a probability distribution of what the actual change was. From the data given we can estimate the chance that new home sales grew in feb. Half the distribution will be above the 4.7% figure. And about 19% will be between 0 and 4.7% (assuming their margin of error is the usual 2 standard deviations, then 4.7% is about half a standard dev).Therefore, the data isn’t MEANINGLESS, it MEANS exactly that there is a 69% chance that new home sales grew in feb and a 31% they didnt.

  6. Mike   March 26, 2009 at 1:13 pm

    Nick Giudici: The graph goes in 200,000 increments and starts at 200,000. The graph would only go up one tick if it started at zero.

  7. Anonymous   March 26, 2009 at 1:48 pm

    Re: Comments by ‘Guest on 2009-03-26 12:59:36′”If a range does not contain zero, the change is statistically significant. If it does contain zero, the change is not statistically significant; that is, it is uncertain whether there was an increase or decrease.” (http://www.census.gov/const/newressales.pdf)Lies, damn lies and statistics!

  8. Mike   March 26, 2009 at 1:52 pm

    Im gona have to go with Mike on this.. def wouldnt make a difernce the angle and speed of decline is whats important. Interesting article, Bravo!

  9. Anonymous   March 26, 2009 at 2:32 pm

    So, your complaint is that the media should be reporting the year over year number instead of the month to month number. Right?Okay. Fair enough. Well, it would be “fair enough” if you were consistent.Here’s how you cover GDP: GDP = -6.2%So here you use the quarterly number. Why not stress the YoY number? It’s right there on the chart you posted.

  10. Sarah   March 26, 2009 at 4:14 pm

    RE: Comments by ‘Anonymous on 2009-03-26 13:48:36 ‘>> Lies, damn lies and statistics!Nothing that Guest said was a lie. He (assuming he’s a dude) even qualified his assumption of a normal distribution (the bell-curve).The Census bureau blurb that you quoted is simply calling anything less than 100% certainty “insignificant”. If a range doesn’t contain zero, then it is either all positive or all negative, and the folks in the Census bureau can say, ‘Sweet, we are CERTAIN that prices increased/decreased. That’s significant!’Guest put it in more concrete terms. The range does contain zero, but rather than just saying ‘Well, we are not certain what happened…’ Guest puts it this way: “…there is a 69% chance that new home sales grew in feb and a 31% they didnt.”That may be “insignificant” by the Census definition, but it is NOT MEANINGLESS. A quantitative statement (with qualified assumptions) carries more meaning than evidently you can see.

  11. Anonymous   March 26, 2009 at 5:09 pm

    Speaking of mathematically innumerate, I have noticed that very often my local newspaper (the Chicago Tribune) will print articles with numbers that very clearly have the wrong number of zeroes, it seems most writers have no grasp of the difference between 100 million and 100 billion.

  12. Guest   March 26, 2009 at 5:41 pm

    make dumb ppl think the economy isn’t as screwed as it actually is….i love media

  13. Guest   March 26, 2009 at 6:26 pm

    Is the distribution of new home sales normal?

  14. Guest   March 26, 2009 at 6:32 pm

    And to really mess things up they “seasonally adjust” the numbers. What does it mean to seasonally adjust numbers after a crash like we’ve had, when we don’t have enough post-crash seasons to do this with? I wonder what the non-seasonally-adjusted numbers are?

  15. Anonymous   March 26, 2009 at 6:41 pm

    Since the median sale price (200,900) is less than the average (251,000). This seems to imply that the distribution is skewed left, but I’m not sure by how much. The price distribution appears to be skewed more left than Feb. of last year, (see http://www.census.gov/const/www/newressalesindex.html US monthly percent distribution).

  16. Guest   March 26, 2009 at 6:42 pm

    This data mis-interpretation is a kissing cousin of the charts that show 1) rate of inflation 2) real GDP growth/decline from pre-1983 to the present. Because the methods of calculation of the rate of inflation has been drastically changed multiple times since 1980 – every time to produce a lower inflation rate number for the same raw data – you get a picture of a “Great Moderation” in which inflation has been all but disappearing (and now is approaching zero, justifying “disinflation” as a rallying cry for negative real interest rates). Because real GDP is calculated using inflation rate numbers the distortion in historical perspective carries through – real GDP appears much higher now than it would using pre-1983 inflation calculation models. So, all comparisons in inflation rates and real GDP levels between the recessions/inflations up to 1983 with the current situation are statistically meaningless.

  17. Guest   March 26, 2009 at 6:57 pm

    Then there’s the birth-death adjustment in the preliminary unemployment numbers. This is a remarkably arbitrary guess (not based on any counting of jobs created) about how many new jobs were created by businesses during the reporting period. For the last year and a half, as the unemployment rate has been going up and the net job creation/loss has been going down, the preliminary birth-death adjustment keeps on guessing that new jobs are being created all over the place every month. This leads to a lower preliminary number for job loss and unemployment rate, which – what a surprise – turns out to be higher when the corrected numbers are released (without the birth-death delusion) the next month. So, every month we get a report of today’s employment numbers and rate of unemployment (preliminary and artificially lowered by the birth-death delusion) which is compared with last month’s corrected employment numbers and rate of unemployment (now higher that originally reported). Like magic, the monthly rate of change (deterioration) is reported to be much less than it would be if they compared preliminary with preliminary or corrected with corrected.

  18. Eric L   March 26, 2009 at 8:05 pm

    Quote: “Here is a life lesson for you: Never believe . . .people who read teleprompters. They have no idea what they are doing, they are reading what someone else wrote.”Now let’s see, who does this bring to mind . . .?

  19. Greg   March 26, 2009 at 9:17 pm

    Your math is correct. You cannot state with 90% confidence that home sales rose in Feb over Jan. However, you can state this with *some* degree of confidence. Since the mean is actually up, you can say that the “best estimate is that home sales rose in Feb over Jan.”. Another way to consider this: you have a 50%/50% chance that home sales are up by at least 4.7%. What are the odds that home sales are up >0%? Probably somewhere around 65% (just approximating, you could actually do the math).

  20. Rich Burke   March 26, 2009 at 11:45 pm

    Thanks for putting he correct info out there. Unlike the Challenger’s O-ring tragedy, when data was overlooked, this is an instance where data is misinterpreted.Makes me wonder if the MSM intentionally tried to mislead folks, or are they just mathematically lazy?Rich Burke, Bellevue, WA

  21. Anonymous   March 27, 2009 at 7:33 am

    Here is a life lesson for you: Never believe news people who read teleprompters.A teleprompter is only as good as the data or words that are input.If some people can`t figure out how to put the right words in the “magic box”, that`s their problem.

  22. Anonymous   March 27, 2009 at 8:46 am

    there are many people looking for Advice Buying New Home, things are starting to look very bleak, I wonder how long it’ll take for things to normalise after the recession ends.

  23. george4title   March 27, 2009 at 8:15 pm

    Barry great report – would love to interview you for my youtube channel over skypethanksgeorge4title

  24. George Harter   March 30, 2009 at 4:31 am

    Gee, as a dumb BioInformatics dude, I am confused about the statistical “Noise” in the the above discussion. If the graph is close to accuracy, Housing Sales suck. Why try to read tea leaves to fire up another Dead Cat rally??? Housing, Real Estate scammers, have been saying the same garbage since last Spring!!GOOD TIMES ARE JUST AROUND THE CORNER, SEE I HAVE STATISTICALLY VALID DATA WHICH INDICATES THAT POSSIBLY, MAYBE THINGS JUST MIGHT BE GETTING BETTER FAST!!! HURRY!! DON’T WAIT, THE TIME TO BUY IS NOW!!!KISS MY A**. Down 41% Y/Y and actually trending down. Any optimists out there are welcome to purchase from me massive quantities of SOUND CLEVELAND REAL ESTATE at a Real Bargain(ca. $2k/home!!!!!). I am also taking bids on the Verrazano Narrows Bridge right now, sealed, the starting is $147,000,000: at auction end I would prefer GOLD at time of settlement. NYPortAuthority will grant possession immediately, I HAVE FRIENDS!! Believe ME, I have nothing to gain by lying!!(Heck, simple linear regression tells me we could see 265,000 units by Sept.) I’d rather build a curve on data instead of on a one month sample. Truthfully, I WILL TAKE FRIENDLY GENTLEMAN’S WAGERS on the 265K Unit sales number. Nobody is AWAKE, the next set of foreclosures will SWAMP US ALL, RUBIN-GEITHNER-BERNANKE-OBAMA may wish not, it is still coming. The Service Companies will do it all on their own!A market for NEW homes????? Sadly, buying foreclosures is no route to wealth either. Rentals???? Resale??? KISS MY A**.George HarterBaghdadontheHudson, USA